Google lost the lawsuit over SERP scraping: analysis of the SerpApi case

On July 20, 2026, the federal court for the Northern District of California dismissed Google's lawsuit against SerpApi — a company that sells access to search results through an API. Google attempted to apply DMCA anti-circumvention provisions to SERP scraping — the same provisions that were written against circumventing DVD protection. The court responded: technical barriers around data that are not protected by copyright are not protected by copyright law. Let's break down what the court actually decided, what it did not decide, and what this means for everyone collecting data from search results.

What Happened: Timeline of the Case

Google filed the lawsuit on December 19, 2025 (case 25-cv-10826-YGR, Northern District of California). The defendant is SerpApi, a service that provides Google Search results in a structured format via an API. A significant portion of the SEO industry relies on its data: position trackers, keyword research tools, competitive intelligence dashboards.

Key facts from the case materials:

SearchGuard — an anti-scraping system launched by Google in January 2025. It sends JavaScript challenges to requests from unidentified sources and requires confirmation that a human is behind the browser. Automation typically fails these checks and gets blocked.
According to Google, SerpApi generated "hundreds of millions of artificial search queries per day," and the volume of requests increased by approximately 25,000% over two years.
Google sought statutory damages — from $200 to $2500 for each act of circumvention. With millions of requests, the total would far exceed SerpApi's annual revenue, which is measured in the low millions of dollars. In other words, a successful lawsuit would effectively shut down the company.
SerpApi filed a motion to dismiss in February 2026. The decision was made by Chief Judge Yvonne Gonzalez Rogers on July 20, 2026.

Why DMCA Did Not Work

Section 1201 of the DMCA prohibits circumventing technical measures that "effectively control access to a work protected by copyright law." Google's logic: SearchGuard is a protective technical measure, SerpApi circumvents it, therefore a violation is evident.

The court dissected the structure piece by piece. Standard search results — URLs, snippets, actual index data, positions — are public facts, not a "work protected by copyright law." Therefore, SearchGuard in this regard "cannot be considered effectively controlling access to a protected work." Claims based on non-copyrightable results were dismissed with prejudice — without the right to refile.

Google argued that the results page is a "mix" of protected and unprotected content, citing licensed images in the Knowledge Panel as an example. Here, the court took a more lenient stance: these claims were dismissed with the right to amend the complaint. However, on the condition that Google must show facts indicating that SearchGuard protects copyrightable elements with the consent of the copyright holders themselves. Simply placing someone else's licensed image on their page and claiming their anti-bot as protection will not suffice. Google has 21 days to file an amended complaint.

SerpApi's blog statement concluded: the court "rejected Google's attempts to extend the DMCA to control access to public pages." The company's CEO, Julien Halégu, describes the decision as a protection of "open access to public data."

Why This Case Is Bigger Than a Dispute Between Two Companies

The context of the last year and a half explains why Google even went to court. Technically, it had already been pressuring the SERP data market:

In September 2025, Google removed the num=100 parameter, which allowed users to retrieve a hundred results in one query. The cost of fully scraping the results for tools increased by approximately tenfold.
When DataForSEO rolled out a workaround that reduced costs by about 80%, Google blocked it within five days.

So, on a technical level, Google was already winning. The lawsuit was an attempt to add a legal lever to this: to turn anti-bot circumvention into an offense with fixed penalties for each request. This is what the court did not allow. If it had passed, any CAPTCHA or JS challenge on a public page would automatically become a "technical protection of a work" — and any public data parser would default to being a DMCA violator.

The Economics of the Issue: Why There Is a War Over SERP Data

Search results are not just "another source of data." They underpin an entire industry: position tracking, competitor analysis, price monitoring, training datasets for AI models. Google finds itself in a contradictory position: the company built its search on mass collection of others' content, yet wants to fully control access to the aggregated results.

A separate layer involves artificial intelligence. Search results have become a quick way for language models to get an up-to-date picture of the internet without their own crawler. This is why in lawsuits over the last year, AI companies increasingly stand alongside classic scraping services. The argument "we're just reading public pages" sounds the same for both an SEO tool and an AI assistant, and courts are just beginning to figure out where the boundary lies between them. The SerpApi case provides a clear answer to at least one question: copyright does not extend to facts simply because they are behind a CAPTCHA.

The decision continues a line familiar from the hiQ Labs v. LinkedIn case: collecting publicly available data does not become a crime simply because the platform owner dislikes it. The difference is that hiQ involved CFAA (unauthorized access), while here another circumvention route — copyright law — has been closed.

What This Decision Does NOT Overturn

It is important not to confuse a victory in one episode with a blanket immunity. What remains in place:

Technical blocks are still in place. SearchGuard operates exactly as it did on July 19. The court's decision does not open access — it merely removes a specific type of legal liability. Scraping results is still technically challenging and expensive.
Contractual claims and ToS. Violating the user agreement is a separate legal area; the DMCA decision does not address it.
Copyright on specific elements. This is the door the court left ajar. Actual results are facts; images in the Knowledge Panel, licensed content, photographs, user reviews — are not. Collecting "everything from the page" and collecting positions and snippets are legally different actions.
Parallel proceedings. In October 2025, Reddit filed a lawsuit in the Southern District of New York against Perplexity AI, SerpApi, Oxylabs, and AWMProxy, accusing them of circumventing protections and collecting data for AI; an amended complaint was filed on February 6, 2026, and the case is currently in the motion to dismiss stage. Additionally, SerpApi is defending itself against a lawsuit from SearchApi in Texas. The legal landscape remains contentious.

And importantly: this is a district court's first-instance decision, not an appellate precedent. Google can amend its complaint regarding licensed content and then appeal. It is premature to assume that the issue is closed forever.

What This Means Practically

If you are collecting data from search results or planning to do so, a reasonable reaction to the decision is not "now we can do anything," but rather clarifying the working rules.

Separate facts and protected content at the parser level. URLs, titles, snippets, positions, the presence of blocks — this is the factual layer, around which the court's position is strongest. Images, embedded licensed materials, full texts of reviews — are risk zones. Technically, this means creating a whitelist of fields in the extractor, rather than "saving all HTML just in case."
Do not confuse legal risk with operational risk. Your main issue when scraping SERP is not the court, but blocks, CAPTCHAs, and IP quality. The market has tightened: after the removal of num=100, the same amount of data costs significantly more per request. We discussed the practical side — how to scrape different blocks and why sessions drop — in the guide on scraping Google SERP and AI Overviews.
Maintain an adequate load. The figure "25,000% increase in requests" was mentioned in the lawsuit for a reason: an aggressive volume becomes an argument against you — both in court and in the eyes of anti-bots. Reasonable intervals, priority requests instead of total scraping, and caching results reduce both types of risks simultaneously.
Keep geo and IP type aligned with the task. Results are personalized by region, and data scraped "from anywhere" is simply inaccurate. For geo-accurate and stable sessions, residential proxies are most often used; for large volumes of similar technical requests, where sensitivity to detection is lower, data center proxies are more cost-effective.
Document what and why you are collecting. An internal document of half a page — sources, fields, frequency, retention period — is inexpensive, and in any dispute, it distinguishes research collection from uncontrolled scraping. Related legal nuances are discussed in the article on legal use of proxies.

Conclusion

The decision from July 20, 2026, is a significant point in the debate over what scraping public results constitutes. The court refused to consider the anti-bot system as a "protection of a work" when it deals with facts rather than creative content, and definitively closed this claim. At the same time, it left Google 21 days to attempt to approach the issue from the side of licensed elements on the page.

The practical implication for the data market is straightforward: legal pressure on the collection of facts from search results has weakened, but technical pressure has not and is not going to. SEO analytics, competitive intelligence, and price monitoring tools continue to operate in the same manner as before: quality access infrastructure, careful load management, and a clear understanding of which specific fields you are extracting from the page.